Yes Bruce, Alpha works well for he purpose you describe. I typed in "sin xy" and got a thorough response with graphics. It is the "knowledge" portion of the engine description I find lacking. If the "knowledge" referred to in the title is curated information at a location dedicated to Alpha, that is perhaps a poor choice as a source for knowledge at a time when High School students write search engines to acquire domain specific knowledge from the Internet. If the knowledge is acquired from the web, then the acquisition technology is not sufficient. Since I must acquire the information required from another source and load it into Alpha to do any calculations, graphing, etc. If Alpha is as powerful as Mathematica it appears to be an attempt at an online, period license version of Mathematica. It does not have the capabilities of Mathematica however, especially if one can program. So one must acquire information from another source, say Google, and then move that data to Alpha or Mathematica to perform any operations on it. If one owns Mathematica, in what way does Alpha or AlphaPro provide additional functionality?
A truly useful system would accept a query using natural language as Wolfram is working on. Then acquire knowledge from a petabyte curated store (a LRU, LFU, LIRS or some hybrid cache), perhaps a distribute database such as Hadoop which ships code rather than data and finally by a web search, something Wolfram is apparently not working on. Google does not perform Mathematical reasoning or logical reasoning.
I foresee a system with which I could paste in a political statement (speech, article) and have arguments extracted, premises evaluated for truth, reasoning (syllogistic and statistical) evaluated and an analysis presented such that the public may discern rhetoric from reason before voting.
Dreaming aside, for Alpha to be a truly "computational knowledge engine" it must have the "knowledge" available for computation (and by computation I mean more than algebra and calculus, but all forms of computation including logical methods embedded in language, statistical and probabilistic arguments, etc.). Currently AlphaPro can tell me the depth of the Challenger deep or the height of the Eiffel Tower but the knowledge available is extremely limited. It is here that I see Alpha falls short of being a "knowledge engine".
This is not a failing since no other engine comes close, rather this is an opportunity to apply this unique front end to a large knowledge base, something for which the method of acquisition is well known, as is the reasoning engine for example look at IBM's Watson:
"Over the course of the project, we have experimented with logistic regression, support vector machines with linear and nonlinear kernels, boosting, single and multilayer neural nets, decision trees, and locally weighted learning; however, we have consistently found better performance using regularized logistic regression, which is the technique used in the final Watson system." - So the "secret sauce" in Watson is regression.
I think that a failure by Wolfram to pursue the knowledge resource aspect of this product and concentrating on the computuation aspect where they already dominate, would be a tragic management failure. As it is, the use of Alpha is restricted to a very small portion of the population when a much larger number are in search of this adjunctive intelligence - real AI they can use.